Grok Not Loading in Clash? Fix Proxy Rules and Timeouts
Identify where Grok stops loading
When Grok does not load through Clash, the symptom alone does not identify the cause. A blank conversation, an endless loading indicator, a sign-in loop, and an explicit network error can point to different parts of the connection path. Traffic may fail before it reaches Clash, be sent to the wrong policy group, time out while connecting to a proxy node, or reach the service but fail during authentication or page loading. Start by recording what happens, then change one setting at a time.
First, check whether the problem affects Grok only or also affects other websites that use the same proxy group. Open a known HTTPS site in the same browser profile, then try Grok in a private window or a second browser. If other sites work but Grok does not, the general proxy connection is probably available, but this does not rule out a domain-specific routing or DNS problem. If several unrelated sites fail, investigate the active Clash mode, local proxy port, and selected node before changing Grok-specific rules.
Next, open the Clash connection log while reloading Grok. In Clash Verge, Clash Verge Rev, or another graphical client, look for the connection view or log panel; the exact menu wording varies by release. Filter or search for domains that appear at the time of the failed request. Record the destination domain, matched rule, selected proxy group, and final node when the client displays them. If no relevant connection appears, the browser may not be using this Clash instance, an existing connection may have been reused, or traffic may be routed through a different VPN or proxy.
Also distinguish a page-load problem from an account or service problem. If the page shell loads but login, conversation history, or message submission fails, note which action fails and whether the browser displays an error code. Avoid repeatedly signing out, clearing all browser data, and changing several proxy settings at once: these actions can remove useful evidence or introduce a separate authentication issue.
Verify Clash mode and the browser’s proxy path
Clash commonly handles traffic through the operating system’s system proxy, a client-managed proxy setting, or TUN mode. System proxy mode is usually limited to applications that honor the operating system’s HTTP or SOCKS proxy settings. TUN mode can capture more application traffic at the network layer, but it depends on permissions, routes, DNS handling, and the client’s implementation. Turning on TUN is not a universal fix, and using system proxy and TUN together can make the actual path harder to diagnose.
For a controlled test, choose one path. On a desktop, enable the Clash system proxy and temporarily disable browser proxy extensions and browser-specific proxy settings. Reload Grok and check the connection log. If the browser only works when an extension is enabled, its traffic may not have been using Clash’s system proxy. If you prefer TUN, verify that the client reports it as active and permitted, and then confirm that the relevant connections appear in the Clash log.
Check the running client’s actual local port rather than assuming it is 7890. A typical configuration may define an HTTP port, a SOCKS port, or one mixed port for both protocols. The operating system proxy must point to a port that the running core is listening on. If the port was changed by a profile, override, or another application, the proxy toggle can appear enabled while the browser is sending requests somewhere else.
mixed-port: 7890
allow-lan: false
mode: rule
log-level: info
This is an illustrative configuration, not a replacement for the active settings managed by your client. Do not paste it over a subscription-generated profile just to test a port. Check the client’s current port settings and runtime status first. If a local proxy test fails, make sure the core has started successfully and that another program is not occupying the configured port. If the browser is set to use a SOCKS port, confirm that it is configured for SOCKS rather than HTTP.
When testing, close and reopen the affected tab or browser after changing proxy settings. Browsers can keep existing connections alive, so a reload may continue using a connection established before the change. Compare a fresh request in the connection log rather than relying only on the page’s appearance.
Check the matched rule and proxy group
In rule mode, Clash evaluates rules from top to bottom. The first matching rule selects its named policy or proxy group; a final catch-all rule such as MATCH handles connections not matched earlier. Grok may use more than one hostname while loading a page, authenticating, or sending requests. Routing only the visible page’s main domain does not guarantee that every related request follows the same route.
Use the connection log to discover the hostnames that are actually requested. Do not start by adding a long list copied from an old guide: service domains and request flows can change, and routing a broad domain such as x.com through a proxy can affect unrelated X features. Instead, reproduce one failed action, note the relevant hostnames, and check which rules matched them. If a request has no domain shown, inspect whether the client reports a destination IP, whether DNS has already resolved the name, and whether sniffing or TUN settings affect the connection metadata.
For a temporary test, you can add a narrow domain rule above broader rules, using the policy group that you intend to test. The following syntax is an example only; replace Proxy with a group name that exists in your configuration and use a domain confirmed in your own log:
rules:
- DOMAIN-SUFFIX,grok.com,Proxy
- MATCH,FINAL
Do not copy this snippet blindly into a complete profile. A rule must be placed in the correct rules section, and the group name must exactly match an existing proxy group. Many profiles already contain their own final rule, so adding a second catch-all or changing the existing one can route much more traffic than intended. If the configuration is subscription-managed, use the client’s supported rule override or merge feature; direct edits may be overwritten on the next update.
Check the selected group itself. A group may be in manual selection mode, choose a node automatically, or contain only nodes that are unavailable. Select one known-working node temporarily, then reload Grok and check the log again. If the log shows the intended rule and group but a different node, inspect the group’s selection and health-check behavior. Restore the normal group selection after testing so the temporary choice does not silently change how other traffic is routed.
Separate node failures from Grok request timeouts
A timeout can occur during DNS lookup, the connection from Clash to the proxy server, the connection from the proxy to the destination, or the service’s response. The wording shown in a browser is often too general to distinguish these stages. Compare the same Grok action on two nodes from different routes if available, and test an unrelated HTTPS site through each node. If one node fails for multiple sites while another works, the node or its route is a stronger suspect than a Grok-specific rule.
Read the Clash log around the failure. Messages about a refused connection, handshake failure, DNS lookup failure, or timeout have different implications. A refused connection often indicates that a remote endpoint or local port is not accepting the connection. A DNS error directs attention to the resolver path. A timeout means that a response did not arrive before the applicable deadline, but by itself does not prove whether the node, destination, or intermediate route is responsible. Keep the exact message and timestamp so you can compare repeated tests.
Do not begin by increasing every timeout value. Browser navigation, proxy connection establishment, DNS, and application requests can each have different time limits. Raising a setting without identifying the phase may make the page wait longer without repairing the route. First compare nodes, confirm that the matched policy is correct, and repeat the test after closing stale connections. If only one node consistently times out for Grok while other sites work, try a different node before editing global timeout settings.
- If Grok and unrelated HTTPS sites fail on the same node, test another node and verify that the client’s core is connected.
- If other sites work on the same node but Grok requests time out, confirm the matched domains and test a different route for those requests.
- If only one browser fails, compare its proxy settings, extensions, and existing connections with a clean browser profile.
- If the page loads but a particular action fails, record the request domain and the precise browser or Clash error instead of changing the entire configuration.
A node’s latency number is not a full application test. It may measure a particular test URL, may be cached, and does not guarantee that the node can reach every destination. Use it as one diagnostic signal alongside real connection-log entries and a repeatable browser test.
Inspect DNS and domain resolution
DNS can affect both rule matching and the address to which a connection is made. Depending on the configuration and core, DNS requests may be handled by Clash, by the operating system, or through a proxy-related resolver path. Browser secure DNS can also use a resolver independently of the system DNS settings. If the browser resolves a domain outside the path expected by Clash, the connection log may show an IP address where you expected a hostname, or a domain-based rule may not match as you intended.
When a Grok request fails, compare the hostname in the browser’s network information, if available, with the connection log. Check whether the client reports a DNS error, an unexpected destination address, or a rule that matched by IP rather than domain. Do not assume that every mismatch requires turning on fake-IP mode or changing DNS servers. DNS modes have different behavior, and a change can affect other applications or break a profile that relies on specific resolver settings.
If you use a browser’s secure DNS feature, temporarily turn it off only as a controlled diagnostic, then retest while watching the Clash log. If the route changes or domain information becomes visible, the browser resolver was part of the difference. Restore the preferred setting after the test and decide whether the Clash DNS configuration should be adjusted. If you change DNS in a subscription profile, make the change through a supported override and preserve a copy of the original configuration.
After a DNS change, allow the previous DNS result and existing browser connections to expire, or restart the browser to make the test repeatable. Some clients also provide a DNS cache or connection cleanup control; use it only when you know which state it clears. Clearing the browser’s entire history is generally unnecessary for a DNS-routing test and can remove saved sign-in state.
Use a controlled recovery sequence
Once you have collected the log entry, troubleshoot in a sequence that changes as little as possible. Begin with the simplest comparison: check whether the same browser can open other HTTPS sites through the same active Clash mode. Then confirm that Grok’s requests appear in the log, identify the matched rule and group, and select one alternative node for a short test. This separates a client-path problem from a route-specific problem without changing DNS or rewriting the profile.
- Confirm the active path. Check whether system proxy or TUN is enabled, verify the actual local port if using a system proxy, and ensure that no second VPN or proxy is taking precedence.
- Capture one failing request. Reload the page or repeat the failed action while the connection log is open. Record the domain, rule, policy group, node, and error message.
- Test one alternate node. Keep the same mode, browser, and rules. If Grok works, investigate the original node or its route; if it does not, continue checking the matching rule and DNS.
- Adjust only a confirmed rule. Add a narrow domain rule only when the log shows that the request is using an unintended policy. Put it before broader rules and verify the policy-group name.
- Test DNS only when indicated. Change one resolver-related setting at a time, repeat the same request, and restore the previous value if the result is unchanged or other sites break.
After a successful test, restore any temporary browser or group settings and retest the normal setup. Confirm that the relevant Grok requests still match the intended rule, while unrelated destinations continue to follow their previous policies. If the page shell works but authentication or a specific feature still fails, keep the observed error separate from the proxy test: it may be a browser-session, account, or service-side issue rather than a general Clash outage.
When asking for help, provide the client and core names, operating system, active mode, whether other sites work, the exact error text, and a redacted connection-log excerpt showing the matched rule and policy result. Remove subscription URLs, tokens, account identifiers, and private server addresses before sharing logs. A small, time-stamped example is more useful than an entire unfiltered log, and it makes it easier to determine whether the next step belongs in the client, routing rules, DNS, or the proxy route.